Wally869/

synthetic-tool-calling-eval

	Synthetic Tool-Calling Eval




	


	Dataset Summary

A synthetic benchmark for evaluating tool-calling (function-calling) behavior in language models: does the model pick the right tool, fill arguments correctly, and correctly decline to call a tool when nothing in the catalog applies. The dataset has three configs:

catalogs — 126 tool-spec rows flattened from 8 tool catalogs (57 distinct tool names; some tool specs are shared between the nested… See the full description on the dataset page: https://huggingface.co/datasets/Wally869/synthetic-tool-calling-eval.